-
Notifications
You must be signed in to change notification settings - Fork 2.9k
/
leaderelection.go
527 lines (465 loc) · 18.9 KB
/
leaderelection.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
/*
Copyright 2015 The Kubernetes Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
// Package leaderelection implements leader election of a set of endpoints.
// It uses an annotation in the endpoints object to store the record of the
// election state. This implementation does not guarantee that only one
// client is acting as a leader (a.k.a. fencing).
//
// A client only acts on timestamps captured locally to infer the state of the
// leader election. The client does not consider timestamps in the leader
// election record to be accurate because these timestamps may not have been
// produced by a local clock. The implemention does not depend on their
// accuracy and only uses their change to indicate that another client has
// renewed the leader lease. Thus the implementation is tolerant to arbitrary
// clock skew, but is not tolerant to arbitrary clock skew rate.
//
// However the level of tolerance to skew rate can be configured by setting
// RenewDeadline and LeaseDuration appropriately. The tolerance expressed as a
// maximum tolerated ratio of time passed on the fastest node to time passed on
// the slowest node can be approximately achieved with a configuration that sets
// the same ratio of LeaseDuration to RenewDeadline. For example if a user wanted
// to tolerate some nodes progressing forward in time twice as fast as other nodes,
// the user could set LeaseDuration to 60 seconds and RenewDeadline to 30 seconds.
//
// While not required, some method of clock synchronization between nodes in the
// cluster is highly recommended. It's important to keep in mind when configuring
// this client that the tolerance to skew rate varies inversely to master
// availability.
//
// Larger clusters often have a more lenient SLA for API latency. This should be
// taken into account when configuring the client. The rate of leader transitions
// should be monitored and RetryPeriod and LeaseDuration should be increased
// until the rate is stable and acceptably low. It's important to keep in mind
// when configuring this client that the tolerance to API latency varies inversely
// to master availability.
//
// DISCLAIMER: this is an alpha API. This library will likely change significantly
// or even be removed entirely in subsequent releases. Depend on this API at
// your own risk.
package leaderelection
import (
"bytes"
"context"
"fmt"
"sync"
"time"
"k8s.io/apimachinery/pkg/api/errors"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/util/runtime"
"k8s.io/apimachinery/pkg/util/wait"
rl "k8s.io/client-go/tools/leaderelection/resourcelock"
"k8s.io/klog/v2"
"k8s.io/utils/clock"
)
const (
JitterFactor = 1.2
)
// NewLeaderElector creates a LeaderElector from a LeaderElectionConfig
func NewLeaderElector(lec LeaderElectionConfig) (*LeaderElector, error) {
if lec.LeaseDuration <= lec.RenewDeadline {
return nil, fmt.Errorf("leaseDuration must be greater than renewDeadline")
}
if lec.RenewDeadline <= time.Duration(JitterFactor*float64(lec.RetryPeriod)) {
return nil, fmt.Errorf("renewDeadline must be greater than retryPeriod*JitterFactor")
}
if lec.LeaseDuration < 1 {
return nil, fmt.Errorf("leaseDuration must be greater than zero")
}
if lec.RenewDeadline < 1 {
return nil, fmt.Errorf("renewDeadline must be greater than zero")
}
if lec.RetryPeriod < 1 {
return nil, fmt.Errorf("retryPeriod must be greater than zero")
}
if lec.Callbacks.OnStartedLeading == nil {
return nil, fmt.Errorf("OnStartedLeading callback must not be nil")
}
if lec.Callbacks.OnStoppedLeading == nil {
return nil, fmt.Errorf("OnStoppedLeading callback must not be nil")
}
if lec.Lock == nil {
return nil, fmt.Errorf("Lock must not be nil.")
}
id := lec.Lock.Identity()
if id == "" {
return nil, fmt.Errorf("Lock identity is empty")
}
le := LeaderElector{
config: lec,
clock: clock.RealClock{},
metrics: globalMetricsFactory.newLeaderMetrics(),
}
le.metrics.leaderOff(le.config.Name)
return &le, nil
}
type LeaderElectionConfig struct {
// Lock is the resource that will be used for locking
Lock rl.Interface
// LeaseDuration is the duration that non-leader candidates will
// wait to force acquire leadership. This is measured against time of
// last observed ack.
//
// A client needs to wait a full LeaseDuration without observing a change to
// the record before it can attempt to take over. When all clients are
// shutdown and a new set of clients are started with different names against
// the same leader record, they must wait the full LeaseDuration before
// attempting to acquire the lease. Thus LeaseDuration should be as short as
// possible (within your tolerance for clock skew rate) to avoid a possible
// long waits in the scenario.
//
// Core clients default this value to 15 seconds.
LeaseDuration time.Duration
// RenewDeadline is the duration that the acting master will retry
// refreshing leadership before giving up.
//
// Core clients default this value to 10 seconds.
RenewDeadline time.Duration
// RetryPeriod is the duration the LeaderElector clients should wait
// between tries of actions.
//
// Core clients default this value to 2 seconds.
RetryPeriod time.Duration
// Callbacks are callbacks that are triggered during certain lifecycle
// events of the LeaderElector
Callbacks LeaderCallbacks
// WatchDog is the associated health checker
// WatchDog may be null if it's not needed/configured.
WatchDog *HealthzAdaptor
// ReleaseOnCancel should be set true if the lock should be released
// when the run context is cancelled. If you set this to true, you must
// ensure all code guarded by this lease has successfully completed
// prior to cancelling the context, or you may have two processes
// simultaneously acting on the critical path.
ReleaseOnCancel bool
// Name is the name of the resource lock for debugging
Name string
// Coordinated will use the Coordinated Leader Election feature
// WARNING: Coordinated leader election is ALPHA.
Coordinated bool
}
// LeaderCallbacks are callbacks that are triggered during certain
// lifecycle events of the LeaderElector. These are invoked asynchronously.
//
// possible future callbacks:
// - OnChallenge()
type LeaderCallbacks struct {
// OnStartedLeading is called when a LeaderElector client starts leading
OnStartedLeading func(context.Context)
// OnStoppedLeading is called when a LeaderElector client stops leading
OnStoppedLeading func()
// OnNewLeader is called when the client observes a leader that is
// not the previously observed leader. This includes the first observed
// leader when the client starts.
OnNewLeader func(identity string)
}
// LeaderElector is a leader election client.
type LeaderElector struct {
config LeaderElectionConfig
// internal bookkeeping
observedRecord rl.LeaderElectionRecord
observedRawRecord []byte
observedTime time.Time
// used to implement OnNewLeader(), may lag slightly from the
// value observedRecord.HolderIdentity if the transition has
// not yet been reported.
reportedLeader string
// clock is wrapper around time to allow for less flaky testing
clock clock.Clock
// used to lock the observedRecord
observedRecordLock sync.Mutex
metrics leaderMetricsAdapter
}
// Run starts the leader election loop. Run will not return
// before leader election loop is stopped by ctx or it has
// stopped holding the leader lease
func (le *LeaderElector) Run(ctx context.Context) {
defer runtime.HandleCrash()
defer le.config.Callbacks.OnStoppedLeading()
if !le.acquire(ctx) {
return // ctx signalled done
}
ctx, cancel := context.WithCancel(ctx)
defer cancel()
go le.config.Callbacks.OnStartedLeading(ctx)
le.renew(ctx)
}
// RunOrDie starts a client with the provided config or panics if the config
// fails to validate. RunOrDie blocks until leader election loop is
// stopped by ctx or it has stopped holding the leader lease
func RunOrDie(ctx context.Context, lec LeaderElectionConfig) {
le, err := NewLeaderElector(lec)
if err != nil {
panic(err)
}
if lec.WatchDog != nil {
lec.WatchDog.SetLeaderElection(le)
}
le.Run(ctx)
}
// GetLeader returns the identity of the last observed leader or returns the empty string if
// no leader has yet been observed.
// This function is for informational purposes. (e.g. monitoring, logs, etc.)
func (le *LeaderElector) GetLeader() string {
return le.getObservedRecord().HolderIdentity
}
// IsLeader returns true if the last observed leader was this client else returns false.
func (le *LeaderElector) IsLeader() bool {
return le.getObservedRecord().HolderIdentity == le.config.Lock.Identity()
}
// acquire loops calling tryAcquireOrRenew and returns true immediately when tryAcquireOrRenew succeeds.
// Returns false if ctx signals done.
func (le *LeaderElector) acquire(ctx context.Context) bool {
ctx, cancel := context.WithCancel(ctx)
defer cancel()
succeeded := false
desc := le.config.Lock.Describe()
klog.Infof("attempting to acquire leader lease %v...", desc)
wait.JitterUntil(func() {
if !le.config.Coordinated {
succeeded = le.tryAcquireOrRenew(ctx)
} else {
succeeded = le.tryCoordinatedRenew(ctx)
}
le.maybeReportTransition()
if !succeeded {
klog.V(4).Infof("failed to acquire lease %v", desc)
return
}
le.config.Lock.RecordEvent("became leader")
le.metrics.leaderOn(le.config.Name)
klog.Infof("successfully acquired lease %v", desc)
cancel()
}, le.config.RetryPeriod, JitterFactor, true, ctx.Done())
return succeeded
}
// renew loops calling tryAcquireOrRenew and returns immediately when tryAcquireOrRenew fails or ctx signals done.
func (le *LeaderElector) renew(ctx context.Context) {
defer le.config.Lock.RecordEvent("stopped leading")
ctx, cancel := context.WithCancel(ctx)
defer cancel()
wait.Until(func() {
timeoutCtx, timeoutCancel := context.WithTimeout(ctx, le.config.RenewDeadline)
defer timeoutCancel()
err := wait.PollImmediateUntil(le.config.RetryPeriod, func() (bool, error) {
if !le.config.Coordinated {
return le.tryAcquireOrRenew(timeoutCtx), nil
} else {
return le.tryCoordinatedRenew(timeoutCtx), nil
}
}, timeoutCtx.Done())
le.maybeReportTransition()
desc := le.config.Lock.Describe()
if err == nil {
klog.V(5).Infof("successfully renewed lease %v", desc)
return
}
le.metrics.leaderOff(le.config.Name)
klog.Infof("failed to renew lease %v: %v", desc, err)
cancel()
}, le.config.RetryPeriod, ctx.Done())
// if we hold the lease, give it up
if le.config.ReleaseOnCancel {
le.release()
}
}
// release attempts to release the leader lease if we have acquired it.
func (le *LeaderElector) release() bool {
if !le.IsLeader() {
return true
}
now := metav1.NewTime(le.clock.Now())
leaderElectionRecord := rl.LeaderElectionRecord{
LeaderTransitions: le.observedRecord.LeaderTransitions,
LeaseDurationSeconds: 1,
RenewTime: now,
AcquireTime: now,
}
timeoutCtx, timeoutCancel := context.WithTimeout(context.Background(), le.config.RenewDeadline)
defer timeoutCancel()
if err := le.config.Lock.Update(timeoutCtx, leaderElectionRecord); err != nil {
klog.Errorf("Failed to release lock: %v", err)
return false
}
le.setObservedRecord(&leaderElectionRecord)
return true
}
// tryCoordinatedRenew checks if it acquired a lease and tries to renew the
// lease if it has already been acquired. Returns true on success else returns
// false.
func (le *LeaderElector) tryCoordinatedRenew(ctx context.Context) bool {
now := metav1.NewTime(le.clock.Now())
leaderElectionRecord := rl.LeaderElectionRecord{
HolderIdentity: le.config.Lock.Identity(),
LeaseDurationSeconds: int(le.config.LeaseDuration / time.Second),
RenewTime: now,
AcquireTime: now,
}
// 1. obtain the electionRecord
oldLeaderElectionRecord, oldLeaderElectionRawRecord, err := le.config.Lock.Get(ctx)
if err != nil {
if !errors.IsNotFound(err) {
klog.Errorf("error retrieving resource lock %v: %v", le.config.Lock.Describe(), err)
return false
}
klog.Infof("lease lock not found: %v", le.config.Lock.Describe())
return false
}
// 2. Record obtained, check the Identity & Time
if !bytes.Equal(le.observedRawRecord, oldLeaderElectionRawRecord) {
le.setObservedRecord(oldLeaderElectionRecord)
le.observedRawRecord = oldLeaderElectionRawRecord
}
hasExpired := le.observedTime.Add(time.Second * time.Duration(oldLeaderElectionRecord.LeaseDurationSeconds)).Before(now.Time)
if hasExpired {
klog.Infof("lock has expired: %v", le.config.Lock.Describe())
return false
}
if !le.IsLeader() {
klog.V(6).Infof("lock is held by %v and has not yet expired: %v", oldLeaderElectionRecord.HolderIdentity, le.config.Lock.Describe())
return false
}
// 2b. If the lease has been marked as "end of term", don't renew it
if le.IsLeader() && oldLeaderElectionRecord.PreferredHolder != "" {
klog.V(4).Infof("lock is marked as 'end of term': %v", le.config.Lock.Describe())
// TODO: Instead of letting lease expire, the holder may deleted it directly
// This will not be compatible with all controllers, so it needs to be opt-in behavior.
// We must ensure all code guarded by this lease has successfully completed
// prior to releasing or there may be two processes
// simultaneously acting on the critical path.
// Usually once this returns false, the process is terminated..
// xref: OnStoppedLeading
return false
}
// 3. We're going to try to update. The leaderElectionRecord is set to it's default
// here. Let's correct it before updating.
if le.IsLeader() {
leaderElectionRecord.AcquireTime = oldLeaderElectionRecord.AcquireTime
leaderElectionRecord.LeaderTransitions = oldLeaderElectionRecord.LeaderTransitions
leaderElectionRecord.Strategy = oldLeaderElectionRecord.Strategy
le.metrics.slowpathExercised(le.config.Name)
} else {
leaderElectionRecord.LeaderTransitions = oldLeaderElectionRecord.LeaderTransitions + 1
}
// update the lock itself
if err = le.config.Lock.Update(ctx, leaderElectionRecord); err != nil {
klog.Errorf("Failed to update lock: %v", err)
return false
}
le.setObservedRecord(&leaderElectionRecord)
return true
}
// tryAcquireOrRenew tries to acquire a leader lease if it is not already acquired,
// else it tries to renew the lease if it has already been acquired. Returns true
// on success else returns false.
func (le *LeaderElector) tryAcquireOrRenew(ctx context.Context) bool {
now := metav1.NewTime(le.clock.Now())
leaderElectionRecord := rl.LeaderElectionRecord{
HolderIdentity: le.config.Lock.Identity(),
LeaseDurationSeconds: int(le.config.LeaseDuration / time.Second),
RenewTime: now,
AcquireTime: now,
}
// 1. fast path for the leader to update optimistically assuming that the record observed
// last time is the current version.
if le.IsLeader() && le.isLeaseValid(now.Time) {
oldObservedRecord := le.getObservedRecord()
leaderElectionRecord.AcquireTime = oldObservedRecord.AcquireTime
leaderElectionRecord.LeaderTransitions = oldObservedRecord.LeaderTransitions
err := le.config.Lock.Update(ctx, leaderElectionRecord)
if err == nil {
le.setObservedRecord(&leaderElectionRecord)
return true
}
klog.Errorf("Failed to update lock optimistically: %v, falling back to slow path", err)
}
// 2. obtain or create the ElectionRecord
oldLeaderElectionRecord, oldLeaderElectionRawRecord, err := le.config.Lock.Get(ctx)
if err != nil {
if !errors.IsNotFound(err) {
klog.Errorf("error retrieving resource lock %v: %v", le.config.Lock.Describe(), err)
return false
}
if err = le.config.Lock.Create(ctx, leaderElectionRecord); err != nil {
klog.Errorf("error initially creating leader election record: %v", err)
return false
}
le.setObservedRecord(&leaderElectionRecord)
return true
}
// 3. Record obtained, check the Identity & Time
if !bytes.Equal(le.observedRawRecord, oldLeaderElectionRawRecord) {
le.setObservedRecord(oldLeaderElectionRecord)
le.observedRawRecord = oldLeaderElectionRawRecord
}
if len(oldLeaderElectionRecord.HolderIdentity) > 0 && le.isLeaseValid(now.Time) && !le.IsLeader() {
klog.V(4).Infof("lock is held by %v and has not yet expired", oldLeaderElectionRecord.HolderIdentity)
return false
}
// 4. We're going to try to update. The leaderElectionRecord is set to it's default
// here. Let's correct it before updating.
if le.IsLeader() {
leaderElectionRecord.AcquireTime = oldLeaderElectionRecord.AcquireTime
leaderElectionRecord.LeaderTransitions = oldLeaderElectionRecord.LeaderTransitions
le.metrics.slowpathExercised(le.config.Name)
} else {
leaderElectionRecord.LeaderTransitions = oldLeaderElectionRecord.LeaderTransitions + 1
}
// update the lock itself
if err = le.config.Lock.Update(ctx, leaderElectionRecord); err != nil {
klog.Errorf("Failed to update lock: %v", err)
return false
}
le.setObservedRecord(&leaderElectionRecord)
return true
}
func (le *LeaderElector) maybeReportTransition() {
if le.observedRecord.HolderIdentity == le.reportedLeader {
return
}
le.reportedLeader = le.observedRecord.HolderIdentity
if le.config.Callbacks.OnNewLeader != nil {
go le.config.Callbacks.OnNewLeader(le.reportedLeader)
}
}
// Check will determine if the current lease is expired by more than timeout.
func (le *LeaderElector) Check(maxTolerableExpiredLease time.Duration) error {
if !le.IsLeader() {
// Currently not concerned with the case that we are hot standby
return nil
}
// If we are more than timeout seconds after the lease duration that is past the timeout
// on the lease renew. Time to start reporting ourselves as unhealthy. We should have
// died but conditions like deadlock can prevent this. (See #70819)
if le.clock.Since(le.observedTime) > le.config.LeaseDuration+maxTolerableExpiredLease {
return fmt.Errorf("failed election to renew leadership on lease %s", le.config.Name)
}
return nil
}
func (le *LeaderElector) isLeaseValid(now time.Time) bool {
return le.observedTime.Add(time.Second * time.Duration(le.getObservedRecord().LeaseDurationSeconds)).After(now)
}
// setObservedRecord will set a new observedRecord and update observedTime to the current time.
// Protect critical sections with lock.
func (le *LeaderElector) setObservedRecord(observedRecord *rl.LeaderElectionRecord) {
le.observedRecordLock.Lock()
defer le.observedRecordLock.Unlock()
le.observedRecord = *observedRecord
le.observedTime = le.clock.Now()
}
// getObservedRecord returns observersRecord.
// Protect critical sections with lock.
func (le *LeaderElector) getObservedRecord() rl.LeaderElectionRecord {
le.observedRecordLock.Lock()
defer le.observedRecordLock.Unlock()
return le.observedRecord
}